Authors:
Blair Helms
Abstract:
This is a proposal for a tool that would analyze the risk of of particular materials to experience a shortage in ability to produce enough for expected demands and for potential for changes in demand, then the tool would analyze the potential and likely durations of given shortages and what the cost of warehousing would be and what the expected return would be when the shortage appears. This will help companies decide what warehousing they should prepare for the shortages, and how much they could profit by doing so. Because the shortages will come and when they do the companies who are prepared can make huge profits compared to those that don’t.
Background:
During the pandemic shortages skyrocketed due to a lack of parts needed for production. The reason this happened was due to an adoption of a zero inventory or just in time supply chains in an attempt to model what a particular car manufacturer had done with their supply chains. However there was a critical difference. This particular car manufacturer didn’t do just in time for all their materials. They did it for the ones they could always have produced just in time and maintained a warehouse supply of more vulnerable materials.
As a result when all the other auto manufacturers were suffering from semi-conductor shortages, this particular car manufacturer did not. On demand supply chains save money by reducing warehousing costs but increase the risk of cost spikes during supply shocks. Currently, organizations have completely failed en masse to mitigate the effect of these supply shocks. This is a solution to that problem that would be much cheaper than existing solutions. The existing solution is to hire an analyst who will look into the details of all supply chains you depend on and do a manual risk analysis. That is costly and is a one time analysis. This would be fast, automated, cheap, and repeatable. All it would require is the gathering of data and a few analysts to fine tune the algorithm and maybe some machine learning to tune it even better.
Description:
This solution requires a lot of data. This data will have to be tracked down and scraped from practically every node on the supply chain in the world and this data will have to be updated nigh constantly. The more companies that use this technology, the easier this will be as it can use the data they automatically use to update it’s databases from their systems. On top of that data, it will also need to know about the estimated risk of strikes, natural disasters, government policy, and other phenomena that can affect each and every supply node. Using this data, an algorithm will have to be generated. There are two routes that can be used for this. One, a manual algorithm will need to be generated and verified by supply chain analysts then tested against past supply chain scenarios. Finally, this algorithms weights could be tweaked by a simple machine learning algorithm. The benefit of this model is that the software is completely understood and could be copied and verified by a real person. Two, generate the algorithm purely through machine learning. The benefit of this is that a machine learning algorithm may catch connections people do not. The detriment is it can be incredibly wrong in incredibly unpredictable ways. After the algorithm analyzes a manufacturers dependent supply chains, it would then spit out the risk factors of each supply chain. (e.g. how likely a shortage is, the expected cost of said shortage, how much warehousing could mitigate that cost and how much the warehousing itself would cost. This would provide a full cost-benefit analysis of warehousing X of Y material.) It’s up to the manufacturer to decide what to do with this information.
TGCS Reference 3086
